How giosem1/hypernegative works
Project Type and Purpose
• This is a Python library/framework (“Hypernegative”) for evaluating Hyperlink Prediction (HLP) and Negative Sampling (NS) models.
• It provides a modular, reproducible evaluation pipeline for research in HLP and NS.
Main Components (Packages)
a. Data & Loader Layer (e.g., hyperlink_prediction/datasets + loader)
b. Packaging & Config (setup.py, pyproject.toml, __init__.py at root)
c. Pipeline Orchestrator (pipelines package)
d. Utilities (utils package)
f. Negative Sampling Module (negative_sampling package)
g. Hyperlink Prediction Module (hyperlink_prediction package)
Relationships and Interactions
• Pipelines orchestrator invokes components from Data & Loader Layer, Negative Sampling Module, Hyperlink Prediction Module, and Utilities in sequence.
• Data & Loader Layer loads raw data and constructs datasets.
• Utilities provide helper functions (e.g., train/test split).
• Negative Sampling Module generates negative examples and scoring.
• Hyperlink Prediction Module trains and evaluates HLP algorithms.
Architectural Patterns & Design Principles
• Modular separation by concern: data ingestion, model logic, pipeline orchestration, and support utilities.
• Common interface / base classes (HyperlinkPredictionBase, NegativeSamplingAlgorithm) enforce consistency.
• Pipeline pattern: an orchestrator drives a sequence of stages.